Abstract
Clin Chem Lab Med 2013, DOI: 10.1515/cclm-2013-0167; Published online 14/05/2013
Laboratory test results are normally interpreted against a demographic reference interval (RI) defining the range of values observed in health. For some analytes a decision limit (DL) is more appropriate, based on an association between a test result value and an increased risk of a particular outcome.
In this retrospective cohort study, patient numbers living and deceased due to any cause at one year postdischarge (n = 39,964) from a regional hospital were determined for five routine laboratory tests (serum potassium, sodium, chloride, urea and creatinine) using the last result prior to discharge. Odds ratios (ORs) were calculated as a function of test value. Risk-based DL intervals were generated from the upper confidence interval (CI) at OR = 1.
The DLs were in broad concordance with hospital RIs: potassium RI 3.5–5.1 to DL 3.4–4.3 mmol/L; sodium RI 136–145 to DL 136–142 mmol/L; chloride RI 94–110 to DL 100–108 mmol/L; urea RI 2.5–7.1 to DL 1.8–7.1 mmol/L; creatinine RI 71–115 to DL 53–97 µmol/L. OR analysis revealed an increased postdischarge one-year mortality risk within 0.8 mmol/L of the potassium upper reference limit. A similar pattern was found for sodium.
While the DLs for the electrolytes remained unchanged when analysed for age quartiles, the elderly population in the study community may limit the applicability of the findings. Potential confounding factors such as correlation of result values and disease associations were not explored.
In conclusion, the authors present an approach to creating DLs for routine tests from Electronic Medical Record databases.